Finch, W. H., & Finch, M. E. H. (2013). Differential item functioning analysis using a multilevel Rasch mixture model: Investigating the impact of disability status and receipt of testing accommodations . Journal of Applied Measurement , 15 (2), 133–151. http://jampress.org

[no doi located]

Accommodations that were provided to students during the national language achievement test included: extended time, read-aloud directions, read-aloud questions, spelling assistance, restate/rephrase, alternate test setting, test proctor scribes student responses, student marks own responses, and other.

National language achievement test data were examined from 5,966 grade 3 students from 156 schools. The test-takers were selected to be representative by gender and ethnicity, as well as by U.S. geographic region and school setting (e.g., urban). Data on students' disability status and accommodation use were also available: 630 students had disabilities, and 483 students received accommodation/s. Students from 12 disability categories participated, predominantly students with intellectual disabilities, visual impairments, learning disabilities, and speech impairments.

The dependent variable was test-takers' scores on 20 language items comprising a national language achievement test.

Test takers' individual response patterns and performance patterns could be represented by five distinct sets, with two distinct sets at the school level. Students in the highest-performing group had the lowest overall disability rate, and students in the lowest-performing group had the highest overall disability rate. Students in the middle three performance level groups had similar overall disability rates. These trends in the association between performance and presence of disabilities was not connected with any specific disability types or categories. Similarly, higher-performing student groups had lower incidence of being provided accommodations, which seems logical, as receiving accommodations is associated with having disabilities. The multilevel Rasch mixture model data also indicated that "schools could be divided into those with relatively higher achievement and those with lower achievement" (p. 149). Further, "the higher achieving schools had lower overall disability rates, as well as lower rates for each particular disability type. Conversely, the lower achieving schools ... had higher percentages of examinees with identified LD and MR, and higher rates of examinees receiving additional time, and having directions read to them" (p. 149). Because of the link between individual and school-level performance patterns, "school context clearly mattered above and beyond an examinee's ability in terms of acquisition of specific language skills" (p. 149). The researchers concluded that, when examining differential item functioning related to disabilities and accommodations use, the multilevel Rasch mixture model demonstrates benefits. An implication of the study reported by the researchers was that "the school an examinee attends can impact their performance on specific items above and beyond their own personal characteristics, and regardless of their ability level" (p. 150).